An improved iterative decoding algorithm for block turbo codes

Since the introduction of the block turbo code (BTC) except, several soft-input/soft-output (SISO) algorithms have been used in order to softly decode product codes. The classical Chase-Pyndiah algorithm seems to be one with the best trade-off between complexity and performance, especially for low error correction capability t (typically 1 or 2) where it is nearly optimal. However, as an algebraic decoding-based algorithm, the lack of codeword diversity is one of its weakness for BTCs with higher error correction capability and/or non binary BTCs. In this paper, we propose an improved iterative decoding algorithm for BTCs. We present a simple sliding encoding-window (SEW) based decoding algorithm which exploits the cyclic and systematic properties of RS and BCH codes. By adding the SEW algorithm to a classical algebraic decoding method, the proposed decoder can easily generate a list of codewords that are close to the decoded codeword. With the codeword diversity, we can compute more reliable soft output necessary in the turbo decoding process, Monte-Carlo simulations of binary and non-binary BTCs are carried out on Gaussian channels. The results show that the algorithm can improve the error performance up to 1.5 dB relative to the conventional Chase-Pyndiah decoder, while the increase in complexity due to the encoding is minor since it is a low-cost process compared to that of algebraic decoding. Compared to the other encoder-based decoding algorithms in the literature, the proposed algorithm has the advantage that there is no requirement to recompute the generator of parity-check matrix by using Gaussian elimination operations, thus a lower computational complexity